Method and apparatus for data screening

Abstract

The embodiment in the invention discloses a method and an apparatus for data screening. The method and the apparatus use client values as factors to screen clients to obtain the maximum benefit of data screening. The method of the embodiment in the invention comprises: generating a ROC graph of client value screening according to an attribute value of the client value and a hitting client value proportion, wherein the X axis of the ROC graph of client value screening is a client retain proportion obtained by ranking the client value attribute value in a size order, and the Y axis is a hitting client value proportion; pointing to a threshold value of the client retain proportion on the X axis of the ROC graph of client value screening, and selecting a client group, wherein the threshold value of the client retain proportion is selected by a user; and screening the client attribute value data in the client attribute graph of the selected client group according to a received instruction of client data screening.

Description

technical field [0001] The invention relates to the field of communication technology, in particular to a data screening method and device. Background technique [0002] In the telecommunications industry, if users who may leave the network can be predicted more accurately based on known customer information, certain measures can be taken to avoid losses caused by users leaving the network. The algorithms of data mining (DataMining) are generally divided into classification, clustering, regression, etc. The loss prediction method generally uses the classification data mining algorithm, and the typical such algorithm is naive Bayesian ( Bayes), decision tree (DT, Decision Tree), support vector machine (SVM, Support Vector Machine), etc. These data mining algorithms generally use historical user data as input, first divide the historical data into training data and test data, and then find out the relationship between the user's existing known information and the possibility...

AI Technical Summary

Problems solved by technology

However, the revenue contribution of each customer to the operator is obviously different, so it cannot accurately reflect the expected income of the operator. Moreover, the operator can only select a group of customers with the highest probability to retain according to the probability ranking, but the actual In fact, operators have a lot of experience in judging customers leaving the network. For example, if the number of complaints is more than 5, the customer is likely to leave the network and can be retained, but if the number of complaints is more than 50, it may be a low-value customer and may not be retained. The experience of making such a judgment cannot be added to the plan of adjusting and retaining the target.

Images